07 · Cost & License Management at Scale

Tableau licensing (Creator/Explorer/Viewer roles, Level 3 Module 2) has real cost implications at enterprise scale. This module covers license optimization and cost governance, with worked arithmetic on a small representative user base.

1. License tiers recap and cost shape

1. Creator: can build and publish new workbooks/data sources — typically the most expensive tier per seat.
2. Explorer: can build from published data sources (can't create new connections) — mid-cost.
3. Viewer: can only view/interact with published content — lowest cost per seat, but usually licensed in larger blocks or bundled pricing.
4. Cost governance's core question: is each person licensed at the lowest tier that covers their actual usage, not the tier that's simply convenient to assign everyone by default?

2. Worked example: right-sizing a 50-person org

1. Suppose a 50-person org has: 5 people who build new data sources and workbooks (need Creator), 15 who build dashboards from the certified Orders source but never create new connections (need Explorer), and 30 who only view published dashboards like the Level 3 Module 10 project (need Viewer).
2. If every one of the 50 is licensed as Creator "to be safe," and Creator costs (hypothetically, for this arithmetic) 3x an Explorer seat and 6x a Viewer seat, the org pays for 50 Creator-equivalents = 300 Viewer- equivalent cost units, versus a right-sized mix of 5 Creator (5×6=30 units) + 15 Explorer (15×3=45 units) + 30 Viewer (30×1=30 units) = 105 units — a drop from 300 to 105, roughly a 65% reduction, just from tier-matching, with no loss of actual capability for anyone.
3. This is the single highest-leverage cost lever in most Tableau deployments: auditing actual usage (who has ever published a new data source vs. who has only ever built from existing ones vs. who has only ever viewed) against assigned license tier.

3. Identifying over-licensed users

1. Server/Cloud admin views (Level 4 Module 4's Admin Views) show each user's actual activity — a Creator-licensed user with zero "published a new data source" events in the last 90 days is a strong candidate for downgrade to Explorer, pending confirmation they don't need it occasionally (e.g. quarterly, not never).
2. Caution: don't downgrade purely on a snapshot — a Creator who published once, 100 days ago, for a legitimate quarterly report, isn't actually over-licensed; combine usage history with a conversation, not just a dashboard number.

4. Extract and storage costs

1. Beyond seat licenses, extract-heavy deployments (Level 4 Module 1's Backgrounder sizing) carry infrastructure/storage costs that scale with the number and size of .hyper extracts — redundant extracts (e.g. the Level 3 Module 8 scenario's 3 near-duplicate Orders copies, each independently refreshed nightly) triple both refresh compute and storage cost for data that should exist once, certified.
2. Cost governance therefore overlaps directly with content governance (Level 4 Module 3): consolidating duplicate sources isn't just a correctness fix, it's a cost optimization, since fewer, well-governed extracts mean less redundant Backgrounder and storage spend.

5. Building a license/cost review cadence

1. Recommended cadence: quarterly review of (a) license tier vs. actual usage per Section 3, (b) duplicate/redundant data source count per Level 4 Module 2's metrics, and (c) extract refresh volume trend (growing faster than user count growth is a signal worth investigating, per Module 4's capacity discussion).

How It Actually Works

1. License tier is enforced at the session-authorization layer, not the content layer: every action a user attempts (opening a workbook, creating a new connection, publishing) checks the caller's assigned role against a fixed capability table on the Server/Cloud identity record — a Viewer's session literally cannot issue a "create data source" request; the client UI doesn't even expose the option, and the server would reject it as unauthorized even if it did. This is why right-sizing (Section 2) is safe: downgrading a Creator who never creates connections to Explorer removes a capability check they were never exercising, not a capability they were silently relying on.
2. Admin Views' per-user activity history (Section 3) is built from the same historical_events audit table referenced in Level 4 Module 4 — each publish, view, or connection-creation action logs an event row with user, action type, and timestamp; "zero publish-new-data-source events in 90 days" is a straightforward WHERE user=X AND action_type='publish_datasource' AND event_time > now()-90d style query, aggregated per user. This is the same audit-log mechanism that supports Level 4 Module 2's CoE metrics and Module 3's deprecation impact analysis — cost governance, content governance, and CoE reporting all ultimately read the same underlying event history rather than three separate systems.
3. Extract storage cost is measurable directly from the .hyper file size on disk (visible per data source in Admin Views' content storage report) multiplied by refresh frequency's compute cost (Backgrounder job duration × frequency, from Level 4 Module 1/4's sizing) — three near-duplicate Orders copies each refreshing nightly cost approximately 3x the Backgrounder minutes and 3x the storage of one consolidated certified copy, which is why Section 4 frames deduplication (Level 4 Module 3's deprecation workflow) as a cost lever with the exact same mechanism as the correctness argument in Level 3 Module 8 — it's the same redundant .hyper files causing both problems simultaneously.
4. The cost-unit arithmetic in Section 2 and the Exercise models a real billing structure (per-tier seat pricing multiplied by headcount at each tier) — the mechanism worth internalizing is that this total is linear in headcount per tier, so the entire savings in both examples comes from moving people between tiers (changing the multiplier applied to their seat), not from reducing headcount or reducing what anyone is able to do; a downgraded Explorer retains full access to every certified source they build from, just without the publish-new- connection capability check passing.

Cheat sheet

Lever	Mechanism	Where covered
Tier right-sizing	Match license to actual usage, not convenience	Section 2
Usage-based downgrade	Admin Views activity history	Section 3
Extract consolidation	Fewer duplicate sources = less refresh/storage cost	Section 4, Level 4 Module 2/3
Quarterly review	Recurring cadence catches drift early	Section 5

Exercise

Using Section 2's cost-unit model, a 20-person team has 8 Creators, 2 Explorers, 10 Viewers. An audit shows 5 of the 8 Creators have never published a new data source in the last 6 months and only build from existing certified sources. Recompute the cost-unit total before and after downgrading those 5 to Explorer, using the ratio Creator=3, Explorer=1, Viewer=0.5 units: Before = 8(3)+2(1)+10(0.5) = 24+2+5 = 31. After downgrading 5 Creators to Explorer = 3(3)+7(1)+10(0.5) = 9+7+5 = 21, roughly a 32% reduction.